M OLECULAR AND C ELLULAR B IOLOGY,Mar.2007,p.1859–1867Vol.27,No.5 0270-7306/07/$08.00?0doi:10.1128/MCB.01395-06
Copyright?2007,American Society for Microbiology.All Rights Reserved.
A MicroRNA Signature of Hypoxia??
Ritu Kulshreshtha,1Manuela Ferracin,2Sylwia E.Wojcik,3Ramiro Garzon,3Hansjuerg Alder,3 Francisco J.Agosto-Perez,3Ramana Davuluri,3Chang-Gong Liu,3Carlo M.Croce,3
Massimo Negrini,2George A.Calin,3and Mircea Ivan1*
Molecular Oncology Research Institute,Tufts-New England Medical Center,Boston,Massachusetts021111;Department of Experimental and Diagnostic Medicine and Interdepartmental Center for Cancer Research,University of Ferrara, Ferrara44100,Italy2;and Comprehensive Cancer Center,Ohio State University,Columbus,Ohio432103
Received28July2006/Returned for modi?cation28August2006/Accepted10December2006
Recent research has identi?ed critical roles for microRNAs in a large number of cellular processes,including
tumorigenic transformation.While signi?cant progress has been made towards understanding the mecha-
nisms of gene regulation by microRNAs,much less is known about factors affecting the expression of these
noncoding transcripts.Here,we demonstrate for the?rst time a functional link between hypoxia,a well-
documented tumor microenvironment factor,and microRNA expression.Microarray-based expression pro?les
revealed that a speci?c spectrum of microRNAs(including miR-23,-24,-26,-27,-103,-107,-181,-210,and-213)
is induced in response to low oxygen,at least some via a hypoxia-inducible-factor-dependent mechanism.Select
members of this group(miR-26,-107,and-210)decrease proapoptotic signaling in a hypoxic environment,
suggesting an impact of these transcripts on tumor formation.Interestingly,the vast majority of hypoxia-
induced microRNAs are also overexpressed in a variety of human tumors.
MicroRNAs represent approximately1%to2%of the eu-karyotic transcriptome and have been shown to play critical roles in the coordination of cell differentiation,proliferation, death,and metabolism(1,3,6,7,17,24)and more recently in tumorigenesis(2,5,6,10,13,20,21).Indeed,a signi?cant percentage of microRNA-encoding genes are located at fragile sites,minimal loss-of-heterozygosity regions,minimal regions of ampli?cation,or common breakpoint regions in cancers. Moreover,global microRNA expression changes have been described to occur in human cancers and in some cases shown to correlate with the clinico-pathological features of the tumor (2,13,29).However,no mechanism has been proposed to date for these pro?le alterations.
Despite this wealth of data,relatively little is known about microRNA regulation and response to microenvironmental factors.One mechanism involves the activation of speci?c sig-nal transduction pathways that in turn promote the transcrip-tion of certain microRNAs.For example,it was reported that the miR-1genes are targets of serum response factor,a con-verging downstream effector for a variety of oncoproteins and growth factors(30).Another transcription factor,the c-myc oncogene product,was also found to activate the expression of a microRNA cluster(25).
Hypoxia is an essential feature of the neoplastic microenvi-ronment.Tumors with widespread low oxygenation tend to exhibit increased invasion and resistance to conventional ther-apy(9).The molecular mechanisms responsible for the hypoxic survival of neoplastic cells are not fully characterized,and a better understanding of this process may lead to novel strate-gies for pharmacological intervention.
Our data indicate that hypoxia leaves a speci?c mark on microRNA pro?les in a variety of cell types,with a critical contribution of the hypoxia-inducible factor(HIF).Moreover, at least a subgroup of these hypoxia-regulated microRNAs (HRMs)seem to play a role in cell survival in a low-oxygen environment.
Finally,by comparing hypoxia-associated microRNA spectra with published data from a large number of tumors(28),we propose that cancer-associated microRNA pro?les exhibit a hypoxic signature.
MATERIALS AND METHODS
Cell culture and growth conditions.Colon(HT29and HCT116)and breast cancer(MCF7and MDA-MB231)cell lines were obtained from Phil Hinds (Tufts-New England Medical Center,Boston,MA).HT29and HCT116cell lines were maintained in Dulbecco’s modi?ed Eagle’s medium supplemented with 10%fetal bovine serum and MCF7,and MDA-MB231cell lines were maintained in RPMI1640medium supplemented with10%fetal bovine serum.Hypoxic conditions were maintained in an InVivo200hypoxia workstation(Biotrace In-ternational,Ruskinn Life Sciences,United Kingdom)with oxygen maintained at 0.2%.Normoxic controls were propagated at37°C and5%CO2.
Plasmids.The mutant versions of HIF-1?and HIF-2?with double proline-to-alanine substitutions have been described previously(14,15).The three-HRE–thymidine kinase(tk)–luciferase reporter is a HIF-responsive construct containing a tandem of hypoxia-responsive elements from the erythropoietin promoter.Both plasmids were provided by William Kaelin(Dana-Farber Cancer Institute,Boston,MA).
MicroRNA microarray analysis.RNA was extracted by using TRIzol(Invitro-gen,Carlsbad,CA),according to the manufacturer’s protocol.RNA was labeled and hybridized on microRNA microarray chips as previously described(20). Brie?y,5?g of RNA from each sample was biotin labeled during reverse transcription using random hexamers.Hybridization was carried out on the second version of our microRNA chip(ArrayExpress accession number A-MEXP-258),which contains381probes for mature and precursor human microRNAs and393probes for mouse microRNAs,together with controls. There are four spot replicates for each probe on the chip.Hybridization signals were detected by biotin binding of a Streptavidin-Alexa647conjugate using a GenePix4000B scanner(Axon Instruments).Images were quanti?ed using the
*Corresponding author.Mailing address:750Washington Street, Box5609,Boston,MA02111.Phone:(617)636-7514.Fax:(617)636-6127.E-mail:mivan@https://www.doczj.com/doc/632413296.html,.
?Supplemental material for this article may be found at http://mcb
https://www.doczj.com/doc/632413296.html,/.
?Published ahead of print on28December2006.
1859 on January 10, 2016 by guest https://www.doczj.com/doc/632413296.html,/ Downloaded from
GenePix Pro 6.0apparatus (Axon Instruments).We used the microRNA no-menclature according to the microRNA Registry (miRBase http://microrna https://www.doczj.com/doc/632413296.html,/sequences/)at the Sanger Institute and the Genome Browser (https://www.doczj.com/doc/632413296.html,).
Analysis of microarray data.Raw data were normalized and analyzed by GeneSpring GX software version 7.3(Agilent Technologies).The GeneSpring software generated a unique value for each microRNA,determining the average of the results from the four spots.Samples were normalized using the on-chip median normalization feature of GeneSpring software;the result for each cell line time course experiment was normalized to that at time zero.MicroRNAs showing an altered expression across the time course were identi?ed using the ?lter of the n -fold-change tool.In detail,a GeneSpring GX ?lter with the n -fold-change tool was used to point out the microRNAs that,for at least two cell lines,were 1.5-fold upregulated at 24h compared to expression levels at time zero.The 1.5-fold-change threshold was chosen on the basis of its use in previ-ously published articles employing these particular types of microarrays.Analysis of variance (ANOVA)with the Benjamini and Hochberg correction for false-positive reduction was used to compare the average values obtained with microRNA probes in all cell lines at 24and 48h with values at time zero.Hierarchical cluster analysis was performed using average linkage and Pearson correlation as a measure of similarity.
Transfections.Plasmid transfections were performed using Lipofectamine 2000(Invitrogen)as per the manufacturer’s protocol.Fresh medium was added after 5h of transfection,and RNA and protein were harvested 24/48h post-transfection.Transfection of mature microRNAs or inhibitors (Ambion,Inc.)was performed using siPORT NeoFX transfection agent (Ambion,Inc.)accord-ing to the manufacturer’s protocol.Transfection complexes were added to cells at a ?nal oligonucleotide concentration of 20nM,and medium was replaced 24h later.
Luciferase assays using reporters based on HRM promoter fragments.We started by amplifying HRM promoter fragments predicted to encompass HIF sites,as shown in Table S1b in the supplemental material.Thus,the genomic region 4.9kb immediately upstream of miR-24-1was ampli?ed using primers 5?ATACTCGAGCTGCTAGGCCATGCGTGTCC3?(forward)and 5?ATTAA GCTTCAAGAGAGAGTTCACCGCGC3?(reverse)(underlining indicates re-striction sites inserted).For miR-181c,a region of approximately 2.0kb imme-diately upstream was PCR ampli?ed using 5?ATAGGTACCCACTCCACAGC CTGAATG3?(forward)and 5?TATAAGCTTGGTGGGGTAGGTGGCAGG GAAC3?(reverse).For miR-26b,a region situated approximately 3kb upstream of the 5?end (and encompassing a cluster of four predicted HIF sites)was PCR ampli?ed using the following primers:5?ATAGCTAGCGAGACAGATGTCC CGCTCCCAG3?(forward)and 5?ATCGCTAGCACGCTCTTGAATGGGAC GG3?(reverse).Due to the size and CG contents,PCR ampli?cation was done using the Herculase II fusion DNA polymerase (Stratagene).The resulting fragments were cloned in the pGL3basic vector (Promega)(miR-24-1and -181c)or pGL3-tk-luciferase vector (from David Fisher,DFCI)(miR-26b).MCF7or HT29cells were cotransfected using Lipofectamine 2000(Invitrogen)with the reporter plasmids and either of the HIF mutants or the empty vector.Luciferase assays were performed 24h posttransfection by following the manufacturer’s protocol (luciferase assay system;Promega).Luciferase activity (measured as relative light units)was scored using a Femtomaster FB12luminometer (Zylux Corporation)and normalized to that of an equal protein concentration in all samples.
Apoptosis assays.Cells were plated in triplicate in a 96-well plate,transfected with microRNA precursors or inhibitors (Ambion,Inc.)as described above,and scored for caspase activation 72h posttransfection using an Apo-ONE homoge-neous caspase-3/7assay kit (Promega)according to manufacturer’s instructions.The protein concentration was determined using Bradford’s method,and caspase activities for all samples were normalized to that of an equal protein amount.The data are presented as means plus standard deviations from three independent experiments performed in triplicate.
Northern hybridization.Total RNA isolation was performed using the TRIzol reagent (Invitrogen)according to the manufacturer’s instructions.RNA samples (20?g each)were run on 15%acrylamide-denaturing (urea)Criterion precast gels (Bio-Rad Laboratories,Hercules,CA)and then transferred onto a Hy-bond-N ?membrane (Amersham Pharmacia Biotech).The hybridization with [?-32P]ATP was performed at 42°C in 7%sodium dodecyl sulfate (SDS)–0.2M Na 2PO 4(pH 7.0)overnight.Membranes were washed at 42°C,twice in 2?SSPE (1?SSPE is 0.18M NaCl,10mM NaH 2PO 4,and 1mM EDTA [pH 7.7])–0.1%SDS and twice with 0.5?SSPE–0.1%SDS.Blots were stripped by boiling them in 0.1%aqueous SDS–0.1?SSC for 10min and were reprobed several times.As a loading control,we used 5S rRNA stained with ethidium bromide or probing for the U6snRNA.
TaqMan RT-PCR.The method was optimized for microRNA,and reagents,primers,and probes were obtained from Applied Biosystems.Human 18S rRNA was used to normalize all RNA samples.Reverse transcriptase (RT)reactions and real-time PCR (PCR)were performed according to manufacturer protocols.RNA concentrations were determined with a NanoDrop apparatus (NanoDrop Technologies,Inc.),and 1nanogram per sample was used for the assays.All RT reaction mixtures,including no-template controls and RT-minus controls,were run in duplicate in an Applied Biosystems 9700thermocycler.Gene expression levels were quanti?ed using the ABI Prism 7900HT sequence detection system (Applied Biosystems).Analysis was performed in duplicate,including with no-template controls.Relative expression was calculated using the comparative cycle threshold method.
ChIP.HT29cells (60%con?uent)were exposed to hypoxia (0.2%O 2)or normoxia (21%O 2)for 24h.Cross-linking was performed using 1%(vol/vol)formaldehyde for 10min,and the reaction was stopped with 125mM glycine for 5min.Cells were washed twice with ice-cold 1?phosphate-buffered saline and collected by scraping them in 1ml of 1?phosphate-buffered saline supple-mented with 1mM phenylmethylsulfonyl ?uoride (PMSF),followed by centri-fugation and lysis in 400?l of buffer (1%SDS,10mM EDTA,50mM Tris-Cl,pH 8.0,protease inhibitor cocktail,and 1mM PMSF).The resulting lysate was incubated on ice for 10min,followed by sonication for DNA shearing to frag-ments between 200bp and 1kb.The supernatant was recovered and diluted in chromatin immunoprecipitation (ChIP)dilution buffer (1%Triton X-100,2mM EDTA,150mM NaCl,20mM Tris-Cl,pH 8.0,protease inhibitor cocktail,1mM PMSF),followed by immunoprecipitation overnight at 4°C with a polyclonal anti-HIF-1?antiserum (ab2185;Abcam)or whole rabbit antiserum (immuno-globulin G [IgG]control).Immunocomplexes were recovered by the addition of a 50%slurry of salmon sperm DNA-protein A-agarose (Upstate)to the samples and sequentially washed for 4min each in buffer I (20mM Tris,pH 8.0,200mM NaCl,0.5%Triton X-100,0.05%deoxycholate,0.5%NP-40,1mM PMSF),buffer II (20mM Tris,pH 8.0,500mM NaCl,0.5%Triton X-100,0.05%deoxycholate,0.5%NP-40,1mM PMSF),buffer III (10mM Tris,pH 8.0,250mM LiCl,1%Triton X-100,0.1%deoxycholate,0.5%NP-40,0.5mM EDTA,1mM PMSF),and buffer IV (10mM Tris,pH 8.0,5mM EDTA).The immuno-precipitated DNA was retrieved from the beads by incubation in elution buffer (10mM Tris,pH 8.0,5mM EDTA,1%SDS)at 65°C for 1h.Cross-linking was reversed using 200mM NaCl at 65°C overnight followed by proteinase K diges-tion at 47°C for 2h.DNA was then puri?ed using a PCR puri?cation kit (QIAGEN),and PCR was performed using primers spaced approximately 15to 250bp apart and encompassing predicted HIF binding sites in miR-210and -26a-2promoters (see below).
5?GGAGCCTTGACGGTTTGACC 3?(forward)and 5?CGAGGACCAGGG TGACAGTG3?(reverse)were used to PCR amplify the miR-210promoter fragment (210-A)containing predicted HIF binding sites located at positions ?1720and ?1822.For the 210-B fragment containing a predicted HIF consen-sus at position ?1166,the following pair was used:5?GGTCGTGAAGGAGC CTCTAAG3?(forward)and 5?GGACCATCCCACTCACTACAG3?(reverse).Finally,the miR-26a-2promoter fragment (26-A),encompassing a predicted HIF binding site at position ?2860,was ampli?ed using 5?CCAAGGACTATG CACATCACC3?(forward)and 5?GGAAAGGCAGTGAGATGGCAC3?(re-verse).The following primers were used to amplify a region in the promoter of miR-130b (negative control,hypoxia nonresponsive):5?GCGAAACCCCAGCT CTACTA3?(forward)and 5?ACACTCTCACTCTGTCGCCC3?(reverse).For HIF site localization and sequences,see Table S1b in the supplemental material.The PCR products were resolved on a 1.5%agarose gel and visualized by ethidium bromide staining.
RESULTS
MicroRNA expression changes in hypoxia.We hypothesized that microRNAs could be involved in hypoxia response and began addressing this hypothesis using a microarray-based screening.We ?rst analyzed a panel of four human cancer cell lines (MDA-MB231,MCF7,HT29,and HCT116),which were subjected to time course exposure to 0.2%oxygen,a concen-tration often present in the hypoxic regions of tumors.RNA was extracted at speci?c time points (0,8,24,and 48h),and microRNA expression pro?ling was done subsequently.Twenty-seven microRNA probes were identi?ed to be at least 1.5-fold upregulated at 24h in at least two cell lines.The ANOVA,with
1860KULSHRESHTHA ET AL.M OL .C ELL .B IOL .
on January 10, 2016 by guest
https://www.doczj.com/doc/632413296.html,/Downloaded from
the Benjamini and Hochberg correction for false-positive re-duction,was employed to compare the microRNA probes av-erage values in all cell lines at 24and 48h versus values at time zero,yielding in all cases signi?cant P values (?0.05)(see Table 1for a summary of levels of induction at 24and 48h in all cell lines versus those at 0h and corresponding P values).A cluster analysis of time points for each cell line is shown in Fig.1.These probes correspond to 23distinct mature microRNAs,which will henceforth be termed HRMs.
We went on to con?rm the microarray data using Northern analysis (Fig.2a for miR-210)or quantitative RT-PCRs for miR-21,-23,-24,-26,-27,-181,-103,-107,-125,-210,and -213(see Fig.2b for a selection;the rest can be found in Fig.S1a in the supplemental material).Priority was given to HRMs that were subjected to subsequent functional analyses.A Northern blot showing a lack of induction of miR-7-3(not a member of the HRM list)served as negative control (see Fig.S1b in the supplemental material).
Evidence for the role of HIFs in HRM regulation.We next addressed the role of HIFs in HRM regulation.HIFs are well-documented master regulators of the hypoxia response and transcription factors,which are stabilized during hypoxia and coordinate the transcription of a wide variety of genes that are critical for survival under conditions of low oxygen (9).In order to investigate whether HRMs exhibit a signi?cant enrichment in predicted HIF binding sites,we performed an in silico search upstream of the genomic sequences that encode all the known microRNAs (1,039sequences,experimentally demonstrated and predicted;https://www.doczj.com/doc/632413296.html,/sequences /ftp/genomes/hsa.gff).
Since only a few microRNA promoters have been identi?ed experimentally,a 6-kb region (5kb upstream and 1kb down-stream region of the 5?end of the annotated microRNA)was designated as a putative promoter sequence.Indeed,most of the experimentally con?rmed transcription factor binding sites are located within the kb ?5region of the transcription start site.Predicted HIF binding sites were analyzed using the MATCH program and the V$HIF1_Q3(GNNK ACGTG CG GNN;boldfacing indicates the core HIF consensus),V$HIF1_Q5(NGT ACGTGC NGB),and V$HIF1_Q6(N RC GTG NGN)position weight matrices from the TRANSFAC database (version 9.1)(23).The position weight matrices de-scribe the position preferences of different nucleotides in the HIF binding site.We scanned regions around the transcription start sites of all the microRNAs from kilobase positions ?5to ?1using the “minFP_good91.prf”pro?le (the pro?le of cutoff values with a minimum number of false-positive predictions)of MATCH,similarly to searches described previously (16,26,27).We then tested whether these consensus sequences are sig-ni?cantly more abundant in the promoters of the 23HRMs (target set)than in the rest of microRNA-ome.Thus,we gen-erated 50,000groups,each consisting of 23promoters ran-domly selected from the 1,039microRNAs and calculated the number of promoters that contain at least one predicted HIF binding site in both target and random sets of promoters.The search was performed separately for the HIF_Q3and HIF_Q5
TABLE 1.Summary of hypoxia-regulated microRNAs a
MicroRNA Symbol MicroRNA expression ratio in indicated cell line
ANOVA P value
HT29
HCT116MCF7
MDA-MB23124h/0h
48h/0h
24h/0h
48h/0h
24h/0h
48h/0h
24h/0h
48h/0h
miR-103-1MIRN103-1 2.33 3.46 1.11 1.71 2.56 1.72 2.38 3.190.00228miR-103-2MIRN103-2 2.31 2.91 1.22 1.70 2.58 2.07 2.6 3.760.00228miR-106a MIRN106A 1.76 2.03 1.01 1.41 2.18 1.26 1.84 1.520.00241miR-107MIRN107 1.94 2.61 1.07 1.75 2.65 1.66 2.76 3.900.00231miR-107MIRN107 2.04 3.26 1.22 1.86 2.72 1.60 2.47 3.370.00228miR-125b-1MIRN125B1 1.83 2.07 1.07 1.52 1.71 1.64 1.51 1.130.00234miR-181a-1MIRN213 2.29 3.02 1.17 1.44 1.61 1.05 1.87 2.240.00398miR-181a-2MIRN181A 2.01 2.73 1.01 1.44 1.87 1.47 2.13 2.840.00241miR-181b-1MIRN181B1 2.17 3.38 1.17 1.96 1.97 1.47 1.97 2.660.00228miR-181b-2MIRN181B2 2.17 3.33 1.04 1.66 1.73 1.28 1.95 1.980.00268miR-181c MIRN181C 2.74 3.33 1.11 1.48 1.26 1.17 1.6 1.750.0083miR-192MIRN192 1.75 2.19 1.07 1.30 2.7 1.63 1.33 1.860.00262miR-195MIRN195 1.45 2.26 1.51 1.51 1.59 1.15 1.17 1.020.00622miR-21MIRN21 2.81 4.00 1.26 2.09 2.86 1.79 2.28 1.340.00242miR-21MIRN21 2.62 3.51 1.34 1.92 2.6 1.66 1.71 1.120.00231miR-210MIRN210 2.92 3.85 2.47 3.73 4.33 3.77 1.450.940.00241miR-213-5p MIRN213 2.14 2.90 1.3 1.10 1.25 1.04 1.64 1.860.00961miR-23a MIRN23A 2.42 3.730.97 2.02 2.5 1.71 2.71 3.410.00231miR-23b MIRN23B 2.7 3.39 1.05 1.82 1.87 1.45 2.76 3.570.00241miR-24-1-3p MIRN24-1 2.44 4.090.98 1.75 2.22 1.30 2.33 1.520.00483miR-26a MIRN26A1 1.91 2.87 1.12 1.99 2.06 1.58 2.21 2.040.00228miR-26a-1MIRN26A1 2.3 3.46 1.14 2.06 2.5 1.88 1.81 1.720.00228miR-26a-2MIRN26A2 1.94 2.62 1.1 1.57 1.37 1.07 1.54 1.660.0046miR-26b MIRN26B 2.21 3.24 1.29 2.04 3.23 2.40 2.32 2.960.00213miR-27a MIRN27A 1.77 2.48 1.13 1.51 2.14 1.65 1.18 1.440.00241miR-30b MIRN30B 2.74 3.96 1.52 2.11 2.38 1.54 1.5 1.400.00234miR-93-1
MIRN93 1.84 1.79 1.02 1.34 2.18 1.54 1.86 1.560.00231
a
MicroRNA expression numbers represent the ratio of each microarray probe expression value at 24or 48h to that at time zero.The P value was obtained by performing an ANOVA statistical comparison between the mean expression levels at 24and 48h versus that at time zero in all cell lines.
V OL .27,2007A MicroRNA SIGNATURE OF HYPOXIA
1861
on January 10, 2016 by guest
https://www.doczj.com/doc/632413296.html,/Downloaded from
consensus sequences.The selection was performed using the function “sample,”which is part of the random module in the Python 2.4programming language (https://www.doczj.com/doc/632413296.html,/),followed by data analysis using the R programming language (https://www.doczj.com/doc/632413296.html,/).The results of the analysis are sum-marized in Fig.3and Table S1a in the supplemental material,indicating a highly signi?cant enrichment of the HIF binding consensuses in the HRM group (P ?0.00294for HIF_Q3;P ?0.011for the HIF_Q5consensus).The search based on the HIF_Q6matrix did not yield a signi?cant P value (not shown),which is not surprising,given the high probability of such a short and degenerate sequence arising by chance very often in the genome.The next step was to obtain experimental evidence for the role of HIF in HRM regulation.First,we transduced consti-tutively stable HIF-1and -2?subunits versus a vector-only control (pcDNA3.1)in HT29and MCF7cell lines under nor-moxia.HIF stabilization was achieved by substituting the two prolines (at positions 564and 402in the case of HIF-1)in the alpha subunits that are subject to oxygen-dependent hydroxy-lation and proteasomal degradation via VHL-dependent ubiq-uitylation (15,16,22).The activity of exogenous HIFs was con?rmed by cotransfection with an HRE-tk-luciferase re-porter (containing three hypoxia response elements)followed by standard luciferase assay (Fig.S2a in the supplemental material).Expression of miR-103,-210,and -213was
measured
FIG.1.Coordinated hypoxic changes of microRNA expression in colon and breast cancer cell lines.Cluster analysis of four cell lines according to the expression of microRNAs upregulated by hypoxia in at least two cell lines.Expression data were normalized to expression at time zero.
1862KULSHRESHTHA ET AL.M OL .C ELL .B IOL .
on January 10, 2016 by guest
https://www.doczj.com/doc/632413296.html,/Downloaded from
following HIF transfection using a modi?ed real-time RT-PCR protocol which revealed a robust and reproducible increase in the expression of all three transcripts (Fig.4).
We then addressed whether HIF has a direct impact on the putative promoter regions of several HRMs (miR-24,-26,-181,and -210).To this end,we ampli?ed 5kb and 2kb of the genomic regions upstream of miR-24-1and miR-181c,respec-tively.We also ampli?ed a 1.2-kb region approximately 3kb upstream of the 5?end of miR-26b (for details,see Materials and Methods).These regions were chosen based on the pre-dicted HIF binding sites identi?ed by in silico analysis;in order to minimize the chance of missing potentially important sites,we used a “less stringent search,”by considering all the
posi-
FIG.2.Con?rmation of HRM induction by hypoxia by Northern analysis or quantitative RT-PCR.(a)Northern blotting con?rmation of miR-210induction under hypoxia (the mature form is indicated).Lanes NOR,6HY,and 30HY show results under normoxia and at 6h and 30h under hypoxia,respectively.An ethidium bromide-stained gel picture is shown as a loading control.(b)Quantitative RT-PCR con-?rmation of HRM induction by 24-hour hypoxia (H)compared with HRM induction for normoxic controls (N).Bars indicate means from two independent experiments.I bars indicate standard
deviations.
FIG.3.Distribution of random 23-microRNA groups (samples)based on HIF1_Q3(a)or HIF1_Q5(b)binding sites.The arrow indicates where the experimental data (HRMs)fall within the random sample population,with the corresponding P value.
V OL .27,2007A MicroRNA SIGNATURE OF HYPOXIA 1863
on January 10, 2016 by guest
https://www.doczj.com/doc/632413296.html,/Downloaded from
tion weight matrices (HIF_Q3,_Q5,and _Q6)(see Table S1b in the supplemental material).The promoter fragments were subcloned into the pGL3-luciferase (for miR-24and -181)or,as an enhancer,into the pGL3-tk-luciferase (for miR-26)vec-tor.HT29cells were cotransfected with these constructs or the control construct pGL3-luc/pGL3-tk-luciferase and constitu-tively active HIF constructs (HIF1P/A or HIF2P/A)or the empty vector pcDNA3.1(PC),followed by incubation under normoxia or hypoxia for 24h.Standard luciferase assays showed that HIF transduction or hypoxia induced a robust activation of all the promoter-luciferase constructs,supporting a direct role of HIF in HRM upregulation (Fig.5).Interest-ingly,HIF-1had a more robust impact on the promoters of miR-24and miR-181c than HIF-2,which was more ef?cient in activating miR-26.This is unlikely to be attributed to differ-ences in transfection ef?ciency,as parallel controls using a three-HRE (Epo)-luciferase reporter cotransfected with the HIF plasmids showed comparable levels of luciferase induc-tion in response to both isoforms (Fig.S2b in the supplemental material).While these differences between HIF-1and HIF-2were not tested at the level of full HRM induction,recent data argue for speci?c targets and biological effects of the two forms (12).
Additionally,for miR-26a-2and miR-210,we con?rmed the dynamic recruitment of HIF to the promoter by ChIP.As shown in Fig.6,the HIF-1?antibody (but not the control IgG antibody)immunoprecipitated the miR-210and miR-26a-2promoter fragments in hypoxic HT29cells but very little in the normoxic controls.A similar assay performed for a region upstream of miR-130b (which is not an HRM),did not reveal measurable recruitment of HIF upon hypoxia exposure.
Roles of select HRMs in cell survival.The impact of HRMs on cell survival was addressed by direct transfection of mature microRNAs or their antisense inhibitors (4)in human cell lines under hypoxic conditions.We transduced miR-24,-26,-107,-210,and -213into MCF7cells and then incubated them under hypoxia for 48h.The proapoptotic response was interrogated by a caspase-3/7assay (Fig.7a).Ef?cient trans-fer and blockade of microRNAs were con?rmed by
North-
FIG.4.Effect of HIF on speci?c microRNA expression.The im-pact of exogenous constitutively active HIF-1?(HIF1P/A)and HIF-2?(HIF2P/A)subunits on miR-103,-210,and -213expression was deter-mined by quantitative RT-PCR in HT29(a)and MCF7(b)cells.The control was the pcDNA3.1empty vector (PC).Bars indicate means from two independent experiments.I bars indicate standard devia-
tions.
FIG.5.Direct effect of HIF in the up-regulation of select HRMs.Relative luciferase activities of HRM promoter reporter constructs in HT29cells.(a)miR-24-1and -181c promoters in a pGL3context;(b)miR-26b fragment in a pGL3-tk context.The constructs were cotrans-fected with constitutively active HIF-1?(HIF1P/A),HIF-2?(HIF2P/A),or the empty vector pcDNA3.1(PC)and incubated under nor-moxia (NOR).The effect of hypoxia (HYP)on the reporter is also shown.Bars indicate means from three independent experiments.I bars indicate standard
deviations.
FIG.6.Direct recruitment of HIF on the miR-210and miR-26promoters under hypoxia.Chromatin was immunoprecipitated from HT29cells using a HIF-1?antibody or an IgG control,and the en-riched genomic fragment was ampli?ed using primers spanning the candidate HREs located at positions ?1720and ?1822(210-A);?1166(210-B)of the miR-210promoter;or ?2860for the miR-26a-2promoter (26-A).A fragment of the miR-130b promoter was used as a negative control.
1864KULSHRESHTHA ET AL.M OL .C ELL .B IOL .
on January 10, 2016 by guest
https://www.doczj.com/doc/632413296.html,/Downloaded from
ern blotting (Fig.7b).The controls employed in these ex-periments were pre-mir miRNA precursor molecules and negative control 1(Ambion,Inc.).
The antisense experiment is of particular importance since it is aimed at blocking the “natural”induction in response to hypoxia.The overexpression of the sense microRNA is antic-ipated to accentuate the effect of the “physiologic”induction of HRMs.Moreover,sense and antisense microRNAs should exhibit opposite effects on cell death compared to the baseline caspase activity associated with a scrambled control microRNA.At least with respect to miR-26,-107,and -210,we reproduc-ibly noticed caspase inhibition in three independent experi-ments,which points towards a decrease in central components of apoptotic signaling.Of note,the mature forms of miR-103and -107are almost identical and their molecular targets and biological effects are predicted to be highly https://www.doczj.com/doc/632413296.html,para-ble effects on caspase activation were elicited in HT29cells under hypoxia using the same microRNAs (not shown).
Correlations between cancer and hypoxia-speci?c microRNA pro?les.Recent investigations have dissected a large number of human neoplasms for microRNA expression (13,21,28)and identi?ed speci?c alterations from normal cells;however,the mechanism and biological impact of these changes remain elusive.In order to address a potential correlation between the pattern of microRNAs altered in solid cancers and under hyp-oxia,we took advantage of the largest genome-wide microarray pro?ling study published to date,including 540tumor samples from six types of solid cancer (breast,lung,colon,stomach,and pancreatic endocrine tumors and prostate carcinomas)and
corresponding normal samples (28).From the 228micro-RNAs,137exhibiting expression values above threshold in at least 90%of samples were retained,which generated a “com-mon signature”of abnormally expressed microRNAs (pre-sented in Fig.2B and supplementary Tables 10and 11in reference 28).Of note,the microarray search for HRMs was performed using the same pro?ling technology.
The vast majority of HRMs identi?ed by our study (Table 2)are also overexpressed in at least some types of tumors,sug-gesting that hypoxia may represent a key contributing “trigger”for microRNA alterations in cancer.
DISCUSSION
Our work identi?es a group of microRNAs (termed HRMs)which are induced in a hypoxic environment.To our knowl-edge,this is the ?rst report establishing a link between a tu-mor-speci?c stress factor and microRNAs and extends the response to low oxygen beyond the “classic”translated genes.We are fully aware that the microarray-based strategy leaves open the possibility that other microRNAs may respond to hypoxia and were simply not detected by the screen.Indeed,with consideration of the well-recognized technical limitations of microarrays,the microRNAs’world is still expanding and it is conceivable that real HRMs were not present on the chip at that stage.
On the other hand,it is possible that some of the microRNAs identi?ed as upregulated by microarray analysis,but not vali-dated independently by quantitative PCR,are false positives.However,given the extremely high con?rmation rate of the candidates tested,it is conceivable that high proportions of the
TABLE 2.Solid correlations of microRNAs with cancers a
MicroRNA
Expression in cancer cells b
hsa-miR-21................Upregulated in general
hsa-miR-23a..............Upregulated in some cancers (pancreas,colon)hsa-miR-23b..............Downregulated in most cancers (upregulated
in pancreas)
hsa-miR-24-1.............Upregulated in general
hsa-miR-26b..............Upregulated in general (except breast)hsa-miR-27a..............Upregulated in general
hsa-miR-30b..............Upregulated in most cancers (downregulated
in lung and breast)
hsa-miR-93-1.............Upregulated in general
hsa-miR-103-2...........Upregulated in general (except breast)hsa-miR-103-1...........Upregulated in general (except breast)hsa-miR-106a............Upregulated in general
hsa-miR-125b-1.........Upregulated in general (except breast)
hsa-miR-181a-2.........Upregulated in some cancers (especially breast)hsa-miR-181c............No evidence for altered expression
hsa-miR-195..............Upregulated in general (supplemental Table 10
in reference 27)
hsa-miR-210..............Upregulated in general
hsa-miR-213-5p.........Upregulated in general (especially breast)hsa-miR-26a..............Upregulated in general (except breast)hsa-miR-181b-1.........Upregulated in general (except breast)hsa-miR-26a-1...........Upregulated in general (except breast)hsa-miR-192..............No evidence for altered expression hsa-miR-107..............Upregulated in general hsa-miR-181b-1.........Upregulated in general
a The vast majority of HRMs (20out of 23)are also overexpressed in tumors.b
See Fig.
2B and supplemental Tables 10and 11in reference 28.
FIG.7.Antiapoptotic effect of select microRNAs under hypoxia.Blockade of miR-26,-107,and -210with antisense inhibitors leads to an increased apoptotic response in three independent experiments (each performed in triplicate).In contrast,an excess of sense microRNAs decreases the apoptotic response.The dotted line represents the apoptotic caspase-3/7baseline activity in response to negative-control microRNA under hypoxia (P,precursor [sense];A,antisense).(b)Northern blot con?rmation of ef?cient transduction or blockade of miR-210in MCF7cells.MCF7cells were transfected with the precur-sor or antisense miR-210or the scramble control (SCR).U6snRNA is shown as the loading control.
V OL .27,2007A MicroRNA SIGNATURE OF HYPOXIA 1865
on January 10, 2016 by guest
https://www.doczj.com/doc/632413296.html,/Downloaded from
remaining candidates indeed respond to oxygen deprivation and represent bona ?de HRMs.
The direct contribution of HIFs to the upregulation of HRMs was dissected for select microRNAs using a combination of lucif-erase-based reporters (containing fragments of microRNA pro-moters)and chromatin immunoprecipitation.A large variety of direct HIF target genes have been reported,and an indirect,microRNA-mediated component could further add to the com-plexity of the molecular response orchestrated by these transcrip-tion factors.Several in silico methods for target gene prediction have been developed and are publicly available,including PicTar (https://www.doczj.com/doc/632413296.html,),TargetScan 3.0.(https://www.doczj.com/doc/632413296.html,/),and miRBase (https://www.doczj.com/doc/632413296.html,/sequences/).For a given microRNA,these utilize different algorithms and ranking criteria (8,18,19)and are known to produce a partially overlap-ping set of candidates,making the search for targets a complex https://www.doczj.com/doc/632413296.html,ing these three programs,we performed in silico searches for HRM targets,which revealed a highly complex spec-trum,including genes involved in apoptosis and proliferation.This could indeed be highly signi?cant for the response to low oxygen,since hypoxia is known to have an impact on both pro-cesses.
We show experimental evidence that caspase activation is inhibited by several HRMs during hypoxia.Interestingly,and in keeping with this,our searches with the above-mentioned programs predict that several HRMs target core components of apoptosis:BAK1(miR-26),BIM (miR-24and -181),BID (miR-23),caspase-7(miR-23),CASP3(miR-30),APAF1(miR-27),and NIX/BNIP3L (miR-23).In the case of miR-26,we demonstrated a direct antiapoptotic effect with oxygen de-pletion,increasing the possibility that BAK1(a proapoptotic protein)is a relevant target.
With regard to miR-210,another HRM with antiapoptotic effect,in silico searches did not reveal candidate targets that are part of the apoptotic machinery.However,a large vari-ety or other genes can in?uence apoptosis in response to speci?c stresses.For example,one putative target revealed by PicTar is neuronal pentraxin 1(11),which has been shown to mediate apoptosis in ischemic neurons,and miR-210could help neutralize such an effect.Whether such a gene could play a role in apoptosis in nonneuronal cells in low oxygen is not known.
Another process known to be affected by hypoxia is prolif-eration,with many cell types undergoing cell cycle arrest dur-ing oxygen deprivation.HRMs could contribute to this process via predicted targets,such as cdc25A (miR-21,miR-103,and miR-107),cyclin D2(miR-26,miR-103),cyclin E1(miR-26),cyclin H (miR23),and cdk6(miR-26,miR-103).Interestingly,and similarly to the case of apoptosis,several cell cycle genes are predicted to be targeted by multiple HRMs,thereby in-creasing the chance of ef?cient downregulation.
Since the hypoxic response described in this paper involves a multitude of microRNAs,it is conceivable that manipulation of any individual HRM could fail to fully capture the pheno-typic impact of this mechanism in low oxygen.The concerted induction of these HRMs could therefore have a much more robust impact on apoptotic/proliferative behavior when oxygen is limiting for extended periods,such as in cancer.One could speculate that differences between HRM induction in various cell types could contribute to a variability in the response to
hypoxia,with important consequences for cancer progression and response to therapy.
Our analysis shows that a surprisingly high proportion of HRMs are overexpressed in human tumors.The alterations of microRNAs in various cancer types is conceivably the sum of a variety of factors (including oncogene signaling,paracrine fac-tors,and pH alterations),but hypoxia could have a signi?cant impact,setting in motion microtranscripts with biological im-pact on survival and/or proliferation.
ACKNOWLEDGMENTS
This work was supported by NIH grant P30DK-34928and an AACR/PanCan career development award to M.I.;a Kimmel Scholar award to G.A.C.;grants from the Italian Ministry of Public Health,the Italian Association for Cancer Research (AIRC),and Comitato Sostenitori Progetto CAN-2006to M.N.;and grants from the National Cancer Institute to C.M.C.M.F.is a recipient of a fellowship from Fondazione Italiana per la Ricerca sul Cancro (FIRC).
We declare that we have no competing ?nancial interests.
REFERENCES
1.Bartel,D.P.2004.MicroRNAs:genomics,biogenesis,mechanism,and func-tion.Cell 116:281–297.
2.Calin,G.A.,M.Ferracin,A.Cimmino,G.Di Leva,M.Shimizu,S.E.Wojcik,M.V.Iorio,R.Visone,N.I.Sever,M.Fabbri,R.Iuliano,T.Palumbo,F.Pichiorri,C.Roldo,R.Garzon,C.Sevignani,L.Rassenti,H.Alder,S.Volinia,C.G.Liu,T.J.Kipps,M.Negrini,and C.M.Croce.2005.A microRNA signature associated with prognosis and progression in chronic lymphocytic leukemia.N.Engl.J.Med.353:1793–1801.
3.Calin,G.A.,C.G.Liu,C.Sevignani,M.Ferracin,N.Felli,C.D.Dumitru,M.Shimizu,A.Cimmino,S.Zupo,M.Dono,M.L.Dell’Aquila,H.Alder,L.Rassenti,T.J.Kipps,F.Bullrich,M.Negrini,and C.M.Croce.200
4.MicroRNA pro?ling reveals distinct signatures in B cell chronic lymphocytic https://www.doczj.com/doc/632413296.html,A 101:11755–11760.
4.Cheng,A.M.,M.W.Byrom,J.Shelton,and L.P.Ford.200
5.Antisense inhibition of human miRNAs and indications for an involvement of miRNA in cell growth and apoptosis.Nucleic Acids Res.33:1290–1297.
5.Croce,C.M.,and G.A.Calin.2005.miRNAs,cancer,and stem cell division.Cell 122:6–7.
6.Farh,K.K.2005.The widespread impact of mammalian MicroRNAs on mRNA repression and evolution.Science 310:1817–1821.
7.Filipowicz,W.,L.Jaskiewicz,F.A.Kolb,and R.S.Pillai.2005.Post-transcriptional gene silencing by siRNAs and miRNAs.Curr.Opin.Struct.Biol.15:331–341.
8.Grif?ths-Jones,S.,R.J.Grocock,S.van Dongen,A.Bateman,and A.J.Enright.2006.miRBase:microRNA sequences,targets and gene nomencla-ture.Nucleic Acids Res.34:D140–D144.
9.Harris,A.L.2002.Hypoxia—a key regulatory factor in tumour growth.Nat.Rev.Cancer 2:38–47.
10.He,L.,J.M.Thomson,M.T.Hemann,E.Hernando-Monge,D.Mu,S.
Goodson,S.Powers,C.Cordon-Cardo,S.W.Lowe,G.J.Hannon,and S.M.Hammond.2005.A microRNA polycistron as a potential human oncogene.Nature 435:828–833.
11.Hossain,M.A.,J.C.Russell,R.O’Brien,and https://www.doczj.com/doc/632413296.html,terra.2004.Neuronal
pentraxin 1:a novel mediator of hypoxic-ischemic injury in neonatal brain.J.Neurosci.24:4187–4196.
12.Hu,C.J.,S.Iyer,A.Sataur,K.L.Covello,L.A.Chodosh,and M.C.Simon.
2006.Differential regulation of the transcriptional activities of hypoxia-in-ducible factor 1alpha (HIF-1?)and HIF-2?in stem cells.Mol.Cell.Biol.26:3514–3526.
13.Iorio,M.V.,M,Ferracin,C.G.Liu,A.Veronese,R.Spizzo,S.Sabbioni,E.
Magri,M.Pedriali,M.Fabbri,M.Campiglio,S.Menard,J.P.Palazzo,A.Rosenberg,P.Musiani,S.Volinia,I.Nenci,G.A.Calin,P.Querzoli,M.Negrini,and C.M.Croce.2005.MicroRNA gene expression deregulation in human breast cancer.Cancer Res.65:7065–7070.
14.Ivan,M.,K.Kondo,H.Yang,W.Kim,J.Valiando,M.Ohh,A.Salic,J.M.
Asara,https://www.doczj.com/doc/632413296.html,ne,and W.G.Kaelin,Jr.2001.HIF-alpha targeted for VHL-mediated destruction by proline hydroxylation:implications for O 2sensing.Science 292:464–468.
15.Jaakkola,P.,D.R.Mole,Y.M.Tian,M.I.Wilson,J.Gielbert,S.J.Gaskell,
A.Kriegsheim,H.F.Hebestreit,M.Mukherji,C.J.Scho?eld,P.H.Maxwell,C.W.Pugh,and P.J.Ratcliffe.2001.Targeting of HIF-alpha to the von Hippel-Lindau ubiquitylation complex by O 2-regulated prolyl hydroxylation.Science 292:468–472.
16.Jin,V.X.,Y.W.Leu,S.Liyanarachchi,H.Sun,M.Fan,K.P.Nephew,T.H.
Huang,and R.V.Davuluri.2004.Identifying estrogen receptor alpha target
1866KULSHRESHTHA ET AL.M OL .C ELL .B IOL .
on January 10, 2016 by guest
https://www.doczj.com/doc/632413296.html,/Downloaded from
genes using integrated computational genomics and chromatin immunopre-cipitation microarray.Nucleic Acids Res.32:6627–6635.
17.Karp,X.,and V.Ambros.2005.Developmental biology.Encountering microRNAs in cell fate signaling.Science 310:1288–1289.
18.Krek,A.,D.Grun,M.N.Poy,R.Wolf,L.Rosenberg,E.J.Epstein,P.MacMenamin,I.da Piedade,K.C.Gunsalus,M.Stoffel,and https://www.doczj.com/doc/632413296.html,binatorial microRNA target predictions.Nat.Genet.37:495–500.19.Lewis,B.P.,I.H.Shih,M.W.Jones-Rhoades,D.P.Bartel,and C.B.Burge.2003.Prediction of mammalian microRNA targets.Cell 115:787–798.
20.
Liu,C.G.,G.A.Calin,B.Meloon,N.Gamliel,C.Sevignani,M.Ferracin,C.D.Dumitru,M.Shimizu,S.Zupo,M.Dono,H.Alder,F.Bullrich,M.Negrini,and C.M.Croce.2004.An oligonucleotide microchip for genome-wide microRNA pro?ling in human and mouse https://www.doczj.com/doc/632413296.html,A 101:9740–9744.
21.
Lu,J.,G.Getz,E.A.Miska,E.Alvarez-Saavedra,https://www.doczj.com/doc/632413296.html,mb,D.Peck,A.Sweet-Cordero,B.L.Ebert,R.H.Mak,A.A.Ferrando,J.R.Downing,T.Jacks,H.R.Horvitz,and T.R.Golub.2005.MicroRNA expression pro?les classify human cancers.Nature 435:834–838.
22.Masson,N.,C.William,P.H.Maxwell,C.W.Pugh,and P.J.Ratcliffe.2001.Independent function of two destruction domains in hypoxia-inducible fac-tor-alpha chains activated by prolyl hydroxylation.EMBO J.20:5197–5206.23.
Matys,V.,E.Fricke,R.Geffers,E.Gossling,M.Haubrock,R.Hehl,K.Hornischer,D.Karas,A.E.Kel,O.V.Kel-Margoulis,D.U.Kloos,https://www.doczj.com/doc/632413296.html,nd,B.Lewicki-Potapov,H.Michael,R.Munch,I.Reuter,S.Rotert,H.Saxel,M.Scheer,S.Thiele,and E.Wingender.2003.TRANSFAC:transcriptional regulation,from patterns to pro?les.Nucleic Acids Res.31:374–378.
24.Miska,E.A.2005.How microRNAs control cell division,differentiation and
death.Curr.Opin.Genet.Dev.15:563–568.
25.O’Donnell,K.A.,E.A.Wentzel,K.I.Zeller,C.V.Dang,and J.T.Mendell.
2005.c-Myc-regulated microRNAs modulate E2F1expression.Nature 435:839–843.
26.Palaniswamy,S.K.,V.X.Jin,H.Sun,and R.V.Davuluri.2005.OMGProm:
a database of orthologous mammalian gene promoters.Bioinformatics 21:835–836.
27.Sun,H.,S.K.Palaniswamy,T.T.Pohar,V.X.Jin,T.H.Huang,and R.V.
Davuluri.2006.MPromDb:an integrated resource for annotation and visu-alization of mammalian gene promoters and ChIP-chip experimental data.Nucleic Acids Res.34:D98–D103.
28.Volinia,S.,G.A.Calin,C.G.Liu,S.Ambs,A.Cimmino,F.Petrocca,R.
Visone,M.Iorio,C.Roldo,M.Ferracin,R.L.Prueitt,N.Yanaihara,https://www.doczj.com/doc/632413296.html,nza,A.Scarpa,A.Vecchione,M.Negrini,C.C.Harris,and C.M.Croce.2006.A microRNA expression signature of human solid tumors de?nes cancer gene https://www.doczj.com/doc/632413296.html,A 103:2257–2261.
29.Yanaihara,N.,N.Caplen,E.Bowman,M.Seike,K.Kumamoto,M.Yi,R.M.
Stephens,A.Okamoto,J.Yokota,T.Tanaka,G.A.Calin,C.G.Liu,C.M.Croce,and C.C.Harris.2006.Unique microRNA molecular pro?les in lung cancer diagnosis and prognosis.Cancer Cell 9:189–198.
30.Zhao,Y.,E.Samal,and D.Srivastava.2005.Serum response factor regulates
a muscle-speci?c microRNA that targets Hand2during cardiogenesis.Na-ture 436:214–220.
V OL .27,2007
A MicroRNA SIGNATURE OF HYPOXIA 1867
on January 10, 2016 by guest
https://www.doczj.com/doc/632413296.html,/Downloaded from